9

NON-LINEAR MODELS
OBJECTIVES:
At the end of this module, you are expected to accomplish the following:

1. Describe the many non-linear models;

2. Utilize the framework for problem-solving;
3. Create and resolve various non-linear models;
4. Use non-linear model theory to solve practical issues.

OVERVIEW

Throughout this module, we'll presume that some of the factors that influence
decisions may be quantified and represented by mathematical functions. We think there
are quantitative elements to many situations. We further assert that analyzing these
measurable elements can produce unique insights that can help in decision-making.
In contrast to probabilistic models, this section focuses on problem structures that
may be theoretically described by so-called classical deterministic models. We
demonstrate that solutions are generated quite readily using calculus when issues are
accurately modeled by such models. The goal of this module is to illustrate how calculus
may be used to solve both constrained and unconstrained classical deterministic models.
We assume that you are familiar with the basic principles of identifying
derivatives because the solution methods require differential calculus.

MATHEMATICAL REPRESENTATION OF RELATIONSHIPS

We examine functional notation before talking about how objective functions are
represented mathematically.

Functional Notation

Definition

A function is a mathematical relationship where the values of one or more

other variables are used to determine the value of a specific variable (Budnick, et al.,
1988)
As an illustration, the relationship
𝑦 = 𝑓 (𝑥 ) (9.1)
implies that variables x and y are related. In more detail, it means that the value of the variable
y depends on or is a function of the value of the variable x. Y is the dependent (criterion)
variable in this situation, whereas x is the independent (predictor) variable. The value of the
algebraic function f at x is also known as f(x). One and only one value of y is connected to each
value of x, however multiple x values might be connected to the same y value. It is possible to
think of a function from set A to set B as a correspondence rule that assigns a distinct element
f(x) to each x in a specified subset of set A.

The Concept of Optimality

The ultimate aim of decision-making is to attempt to positively affect the objective
function. The state of optimization is when the selected measure of effectiveness is either
minimized or maximized. For instance, reducing the average wait time between a patient's
arrival at the hospital and the arrival of medical personnel might be the goal in the emergency
room of a hospital.

Figure 9.1 – Minima and Maxima (Lifted from Budnick, et al., 1988)

Take a look at Figure 9.1, where one independent variable influences the criteria
variable. Although this continuous objective function has an odd appearance, it can be used to
illustrate several optimality ideas. Suppose that the independent variable x has values that
range from 0 to xm.
A function is said to reach a relative or local maximum at the point x = a if f(a) is greater
than the value of f(x) for any nearby value of x, where variable x denotes the independent
variable. Similar to this, if f(a) is less than f(x) for any nearby value of x, then the function is
said to have reached a relative or local minimum at the point x = a. For instance, in Figure 9.1,
relative maxima occur at locations A and C as well as when x = xm on the curve (note that we
are focusing on continuous functions only). The value of the criterion variable is less than f(xa)
if you choose a position near xa. The same justification is true for sites around x = xc. One
might come to the conclusion that relative minima happen at sites B and D and when x = 0 after
reading the definition of a relative minimum. The value of the criterion variable is bigger than
f(xb) or f(xd), respectively, for independent variable values close to either x = xb or x = xd.

Functional Relationship Estimated

There are numerous circumstances where the precise logical relationships between
variables are unknown. As a result, it is more challenging to express an objective function
theoretically. In these circumstances, we frequently gather pertinent data through sampling to
ascertain whether any discernible regular pattern of behavior is present. If a pattern does appear
to exist, a curve-fitting technique may be used to describe the relationship statistically. The
methods for fitting the curve might vary from "eyeballing" it to more precise methods like
regression analysis.

Problem-Solving Framework
We will apply the framework below to resolve non-linear models, with examples
provided to show how it is used (Budnick, et al., 1988).

1. Verbal Statement of Problem

2. Decisions
a. Verbal Statement
b. Mathematical Definition
3. Criteria
a. Verbal Statement
b. Mathematical Definition
4. Constraining Conditions
a. Verbal Statement
b. Mathematical Statement
5. Mathematical Model
6. Model Outcomes
7. Interpretation of Results

SINGLE-VARIABLE UNCONSTRAINED OPTIMIZATION

Illustration Problem:
The Massachusetts Bay Transit Authority (MBTA) wishes to smooth its demand for
subway services. The subway system operates at capacity during the early morning and late
afternoon rush hours. The MBTA wants to stimulate greater demand during off-peak hours
and to encourage riders who do not need to travel during peak hours to travel during peak
hours to travel at other times. Homemakers shopping in the city during the day are examples
of this category of riders.
The MBTA has been experimenting with off-peak fares in order to determine the effects
on demand. Currently, a program allows passengers to travel for $0.30 rather than the normal
$0.75 between the hours of 10 am and 1 pm. Data have been gathered showing demand under
other fares.
The data, when graphed, offer strong evidence of a linear demand function during off-peak
hours. Since this demand function was not available initially, it is necessary to approximate it
from collected data. Fitting a least-squares regression line to the data, we estimate the demand
equation as
1
𝑞 = 45 − 𝑝
2
Or
𝑝 = −2𝑞 + 90
Where p = off-peak fare (cents), and q = off-peak demand (thousand of riders).
If the objective is to maximize the total revenue during off-peak period, determine the
price to charge per ride.
Applying the Problem-Solving Framework:

1. Verbal Statement of the Problem

Determine the price to charge per ride in order to maximize total revenue during
off-peak daylight hours.
2. Decisions
a. Verbal Statement
Fare to charge for each ride
b. Mathematical Definitions
p = off-peak fare (cents)
3. Criteria
a. Verbal Statement
Total Revenue
b. Mathematical Definitions
𝑇𝑅 = 𝑝𝑞
4. Mathematical Definitions
𝑇𝑅 = 𝑝𝑞
1
𝑇𝑅 = 45𝑝 − 𝑝2
2
5. Constraints
None
6. Model
With no constraints, the model for this problem consists of the objective
function
1
𝑇𝑅 = 𝑓 (𝑝) = 45𝑝 − 𝑝2
2
7. Model Outcomes
The above equation is quadratic in form and graphs as a parabola, concave
downward. To determine whether any relative maxima or minima exist, stationary
points are first identified. Stationary points are those whose tangent slope equals
zero. Taking the first derivative of the revenue function with respect to the variable p,
1
𝑓′(𝑝) = 45 − (2)𝑝
2
𝑓′(𝑝) = 45 − 𝑝
Setting this equal to zero in order to identify points of zero slope, we get
𝑓 ′(𝑝) = 45 − 𝑝 = 0
𝑝∗ = 45
Testing for relative maximum, we take the second derivative of the TR:
𝑓 ′′ (𝑝) = −1
Thus, a relative maximum exists for 𝑓 (𝑝) when p = 45.
7. Interpretation of Results
Since the value of the second derivative is negative throughout the range of p,
we conclude that there exists an absolute maximum for total revenue when a fare of
45 cents is charged.

TWO-VARIABLE UNCONSTRAINED OPTIMIZATION

Illustration Problem:
A refuse incinerator is being planned to service four metropolitan areas. The relative
locations of the metropolitan areas are given by the coordinates, where coordinates are
expressed in kilometers.
 A (10, 30), B (30, 20), C (-20, -20), D (-30, 20)
Environmental management officials wish to determine to incinerator location that
minimizes the sum of the squares of the distances from the incinerator to each metropolitan
area.

Applying the Problem-Solving Framework:

1. Verbal Statement of the Problem

Determine to incinerator location that minimizes the sum of the squares of the
distances from the incinerator to each metropolitan area.
2. Decisions
a. Verbal Statement
Location of proposed incinerator
b. Mathematical Definitions
Coordinates (x, y)
3. Criteria
a. Verbal Statement
Sum of squares of distances from the incinerator to each metropolitan area.
b. Mathematical Definitions
To illustrate, the distance separating point A and the incinerator (x, y) is
𝑑𝐴 = √(𝑥 − 10)2 + (𝑦 − 30)2 (From Distance Formula)
• Since we wish to minimize the sum of the squares of the distances separating
the four areas and the incinerator, our criterion function is
𝑆 = 𝑑𝐴2 + 𝑑𝐵 2 + 𝑑𝐶 2 + 𝑑𝐷 2
𝑆 = 𝑓 (𝑥, 𝑦) = [(𝑥 − 10)2 + (𝑦 − 30)2 ] + [(𝑥 − 30)2 + (𝑦 − 20)2 ]
+ [(𝑥 + 20)2 + (𝑦 + 20)2 ] + [(𝑥 + 30)2 + (𝑦 − 20)2 ]
4. Constraints
None
5. Model
The model for this problem consists of the criterion function
𝑆 = 𝑓 (𝑥, 𝑦) = (𝑥 − 10)2 + (𝑦 − 30)2 + (𝑥 − 30)2 + (𝑦 − 20)2 + (𝑥 + 20)2 + (𝑦 + 20)2
+ (𝑥 + 30)2 + (𝑦 − 20)2
6. Model Outcomes
To determine the values of x and y that minimize S, we must first find the partial
derivatives.
𝑓𝑥 = 2(𝑥 − 10)(1) + 2(𝑥 − 30)(1) + 2(𝑥 + 20)(1) + 2(𝑥 + 30)(1) = 8𝑥 + 20
𝑓𝑦 = 2(𝑦 − 30)(1) + 2(𝑦 − 20)(1) + 2(𝑦 + 20)(1) + 2(𝑦 − 20)(1) = 8𝑦 − 100
If the two partial derivatives are set equal to zero, a stationary point occurs on the graph
of 𝑓 (𝑥, 𝑦) when 𝑥 ∗ = −2.5 and 𝑦 ∗ = 12.5
The second partial derivatives are
𝑓𝑥𝑥 = 8 𝑓𝑥𝑦 = 0
𝑓𝑦𝑦 = 8 𝑓𝑦𝑥 = 0
To test the nature of the stationary point we compute
𝐷(−2.5,12.5) = (8)(8) − (0)2 = 64 > 0
Since D > 0 and both 𝑓𝑥𝑥 and 𝑓𝑦𝑦 are positive, we can conclude that a relative minimum
occurs on 𝑓 (𝑥, 𝑦) when 𝑥 ∗ = −2.5 and 𝑦 ∗ = 12.5
7. Interpretation of Results
 The model outcome suggests that the sum of the squares of the distances from
the incinerator to each metropolitan area will be minimized if the incinerator is located
at coordinates (-2.5, 12.5).

OPTIMIZATION WITH CONSTRAINTS

Illustration Problem:
An automobile manufacturer is considering the installation of two complementary anti-
pollution devices (catalytic converters) in a new model automobile. Configurations of the two
devices have been tested, in terms of installation cost. Exhaust quality tests and cost analyses
uncovered a very strong relationship between each device. A least-squares fit to the data
resulted in the criterion function
𝑅(𝑥1 , 𝑥2 ) = −3𝑥1 2 − 2𝑥2 2 + 20𝑥1 𝑥2
Where 𝑅 (𝑥1 , 𝑥2 ) =reduction is pollution particulates per cubic foot exhaust; 𝑥1 =installation
expenditure for the first converter (dollars); and 𝑥2 = installation expenditure for the second
converter (dollars).
Prior analysis has established that any installation configuration will satisfy federal
and state anti-pollution guidelines. An edict from “above,” however, has established that
exactly $100 is to be spent on the installation of either or both converters. At first, management
was inclined to minimize installation expenditures subject to the satisfaction of pollution levels.

Applying the Problem-Solving Framework:

1. Verbal Statement of the Problem

Determine the cost to be expended for the installation of the catalytic converters that
will maximize the reduction in pollution particulates per cubic foot of exhaust
2. Decisions
a. Verbal Statement
Installation expenditure for the converters (dollars)
b. Mathematical Definitions
𝑥1 =Installation expenditure for the first converter (dollars)
𝑥2 =Installation expenditure for the second converter (dollars)
3. Criteria
a. Verbal Statement
Reduction in pollution particulates per cubic foot of exhaust
b. Mathematical Definitions
𝑅(𝑥1 , 𝑥2 ) = −3𝑥1 2 − 2𝑥2 2 + 20𝑥1 𝑥2

4. Constraints
𝑥1 + 𝑥2 = 100
5. Model
𝑅(𝑥1 , 𝑥2 ) = −3𝑥1 2 − 2𝑥2 2 + 20𝑥1 𝑥2
Subject to:
𝑥1 + 𝑥2 = 100
Solving the constraint for 𝑥1 and substituting this in the objective function, we reduce
the problem to maximizing
𝑅(𝑥2 ) = −25𝑥2 2 + 2600𝑥2 − 30,000
Hence, the original two-variable and one-constraint problem has been transformed to a
problem stated in term of one independent variable.
6. Model Outcomes
𝑅(𝑥2 ) = −25𝑥2 2 + 2600𝑥2 − 30,000
 𝑅′𝑥2 = −50𝑥2 + 2,600 = 0
Or
𝑥2 ∗ = 52
Testing the stationary point, we get
𝑅′′(𝑥2 ) = −50
Which confirms that R is maximized when 𝑥2 ∗ = 52.
7. Interpretation of Results
Thus, $48 and $52, respectively, should be expended for the installation of the
two catalytic converters. This results in a reduction of 37,600 pollution particulates per
cubic foot of exhaust.

CONCEPT REVIEW

This module has introduced the mathematical representation of functional

relationships. Not only have we presented the notation of mathematical functions; we
have also attempted to give you a sense of how analysts formulate these relationships.
This module has also introduced the notion of optimality.
Calculus-based optimization methods have been a primary focus in this part. We
have begun to illustrate the diversity of applications, first with unconstrained optimization
models and with constrained optimization models
NAME: ____________________________________ SCORE: __________________

COURSE/YEAR: _______________________________ DATE: _______________

EXERCISE 8

1. A pervasive problem among organizations is determining the appropriate quantities of inventory to

keep on hand. A common decision relates to the quantities to order each time the firm replenishes
its inventory (if the inventoried item is supplied by a vendor). A firm has determined the following
cost function, which expresses the annual cost of purchasing, owning, and maintaining its inventory
as a function of the size of each other (q),
250,000
𝑐= + 400𝑞 + 50,000
𝑞
a. Determine the order quantity which results in minimum annual inventory cost.
b. What is the minimum cost?
c. Confirm the nature of the stationary point.

 SOLUTION:
 NAME: ____________________________________ SCORE: __________________

COURSE/YEAR: _______________________________ DATE: _______________

2. A firm sells two products for which it has determined the demand functions 𝑞1 = 110 − 4𝑝1 − 𝑝2
and 𝑞2 = 90 − 2𝑝1 − 3𝑝2 , where 𝑞𝑖 =demand (in units per day) for the ith product and 𝑝𝑖 =price
(dollars per unit) for the ith product. Note that the quantity demanded of one product depends on
the price of the product as well as the price of the other product.
Determine the prices that should be charged for each product in order to maximize total revenue per
day from the sale of the two products.

 SOLUTION:
NAME: ____________________________________ SCORE: __________________

COURSE/YEAR: _______________________________ DATE: _______________

3. A community is planning to build a solid waste treatment facility. A major component of the facility
is a solid waste agitation pool. The pool is to be circular in shape and is to have a volume capacity
of 500,000 cubic feet. Engineers estimate the construction costs as a function of the surface area of
the base and wall of the pool. Costs are expected to be $20 per square foot for the base of the pool
and $15 per square foot of wall surface.
a. Determine the dimensions of r and h (both in feet) that provide a capacity of 500,000 cubic feet
at minimum costs.
b. What are the minimum construction costs?

 SOLUTION:
 NAME: ____________________________________ SCORE: __________________

COURSE/YEAR: _______________________________ DATE: _______________

4. A rectangular container with top is to have a volume of 27,000 cubic centimeters. The objective is
to minimize the amount of material used in constructing container. Thus, the surface area is to be
minimized. Given the sketch in the figure below, determine the dimensions that minimize the
surface area.

x
y

 SOLUTION: